The PSMOOTH Procedure

In the search for complex disease genes, linkage and/or association tests are often performed on markers from a genome-wide scan or SNPs from a finely scaled map. This means hundreds or even thousands of hypotheses are being simultaneously tested. Plotting the negative log -values of all the marker tests will reveal many peaks that indicate significant test results, some of which are false positives. In order to reduce the number of false positives or improve power, smoothing methods can be applied that take into account -values from neighboring, and possibly correlated, markers. That is, the peak length can be used to indicate significance in addition to the peak height. The PSMOOTH procedure offers smoothing methods that implement Simes’ method (1986), Fisher’s method (1932), and/or the truncated product method (TPM) (Zaykin et al. 2002) for multiple hypothesis testing. These methods modify the -value from each marker test by using a function of its original -value and the -values of the tests on the nearest markers. Since the number of hypothesis tests being performed is not reduced, adjustments to correct the smoothed -values for multiple testing are available as well.

PROC PSMOOTH can take any data set containing any number of columns of -values as an input data set, including the output data sets from the CASECONTROL and FAMILY procedures (see Chapter 5 and Chapter 6 for more information).